Research Enginner @googledeepmind in🇨🇭. Prev @EPFL, @Google, @AmazonScience and @ecentua
Zurich, Switzerland
Joined March 2016
- Tweets126
- Following337
- Followers268
- Likes1.4K
Pinned Tweet
Happy to share our latest preprint with @apopescubelis : "Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation"
📝preprint: arxiv.org/abs/2401.06688
Giorgos Vernikos retweeted
Congrats @redpony and team! This is an amazing achievement and shows how AI research can power applications for social good! 🥳
Also it was personally fun sitting close by in the office and watching this project take off!
I'm so happy to be able to announce the first general purpose sign-language-to-text translation (SL2T) model from my team that's out today and powering a new ASL input feature on @Android phones (and don't worry, we're not stopping at ASL).
Nelly is amazing! I can’t recommend working with her enough !!!
Are you interested to join @GoogleDeepMind as a student researcher (PhD students)?
I am hiring for a fun project on reward modeling for world models!
If you are interested on video understanding/generation and reward models/critiquing, get in touch and apply below! 👇
1
103
Giorgos Vernikos retweeted
If you’re at @iclr_conf this week, come check out our spotlight poster INCLUDE during the Thursday 3:00–5:30pm session!
I will be there to chat about all things multilingual & multicultural evaluation.
Feel free to reach out anytime during the conference. I’d love to connect!
🚀 Introducing INCLUDE 🌍: A multilingual LLM evaluation benchmark spanning 44 languages!
Contains *newly-collected* data, prioritizing *regional knowledge*.
Setting the stage for truly global AI evaluation.
Ready to see how your model measures up?
#AI #Multilingual #LLM #NLProc
Giorgos Vernikos retweeted
Disappointed with #ICLR or #NAACL reviews? Consider submitting your work at #Repl4NLP @naaclmeeting , whether it's full papers, extended abstracts, or cross-submissions. 🔥
Details on submissions 👉 sites.google.com/view/repl4n…
⏰ Deadline January 30
Giorgos Vernikos retweeted
📢📢 Call for papers
#Repl4NLP @naaclmeeting
Consider submitting your work -full papers, extended abstracts, or cross-submissions!
✨ Direct paper submission deadline: Jan 30, 2025
✨ ARR commitment deadline: Feb 20, 2025
More details on our website: sites.google.com/view/repl4n…
Giorgos Vernikos retweeted
We are organizing Repl4NLP 2025 along with @fredahshi @gvernikos @vaibhav_adlakha @xiang_lorraine @mbodhisattwa. The workshop will be co-located with NAACL 2025 in Albuquerque, New Mexico and we plan to have a great panel of speakers.
Consider submitting your coolest work!
📢📢 Call for papers
#Repl4NLP @naaclmeeting
Consider submitting your work -full papers, extended abstracts, or cross-submissions!
✨ Direct paper submission deadline: Jan 30, 2025
✨ ARR commitment deadline: Feb 20, 2025
More details on our website: sites.google.com/view/repl4n…
Giorgos Vernikos retweeted
Paper #2: Language and Task Arithmetic with Parameter-Efficient Layers for Zero-Shot Summarization
Link: aclanthology.org/2024.mrl-1.…
Ever wondered how we can do LLM weight arithmetic to enable models to handle tasks in languages in zero shot style? The authors have a solution.
Giorgos Vernikos retweeted
🚀 We've released the Lite version of Global-MMLU!
✅ Faster, easier evaluation
✅ Balanced culturally sensitive and culturally agnostic knowledge
Try it at 🤗: huggingface.co/datasets/Cohe…
We have released Global-MMLU-lite 🔥
This is designed to run more efficiently while giving a good estimate of overall performance -- which balances both culturally sensitive and culturally agnostic examples. 🎉
huggingface.co/datasets/Cohe…
Giorgos Vernikos retweeted
Introducing Global-MMLU🌍: A multilingual benchmark featuring MMLU translations in 42 languages crafted with:
✅ Human curation
✅ Extensive metadata
✅ Insights into cultural sensitivity
Proud to have collaborated with @singhshiviii and @CohereForAI to bring this work to life!
Giorgos Vernikos retweeted
🚀 Introducing INCLUDE 🌍: A multilingual LLM evaluation benchmark spanning 44 languages!
Contains *newly-collected* data, prioritizing *regional knowledge*.
Setting the stage for truly global AI evaluation.
Ready to see how your model measures up?
#AI #Multilingual #LLM #NLProc
Giorgos Vernikos retweeted
Super excited - we received ✨✨best paper award ✨✨ for our paper at the Multilingual Representation Learning workshop #EMNLP2024!!! This was a fun internship project thanks to @seb_ruder @priyanka_17 and @cindyxinyiwang @PfeiffJo @maynez_joshua!
Can we achieve effective zero-shot summarization by combining language & task information from existing PEFT adapters with weight arithmetic?
Our paper, accepted at the @mrl2024_emnlp workshop #EMNLP2024, explores this question! A 🧵:
ALT Language and Task Arithmetic with Parameter-Efficient Layers for Zero-Shot Summarization (arxiv.org/pdf/2311.09344)
Giorgos Vernikos retweeted
I am in Miami 🌴 attending #EMNLP2024! Tomorrow I will be presenting my @GoogleDeepMind internship paper at the @mrl2024_emnlp workshop at 11:00am (poster session) and as an oral at 2:30pm! If you are still around and interested in multilingual model merging, stop by!🎉
Can we achieve effective zero-shot summarization by combining language & task information from existing PEFT adapters with weight arithmetic?
Our paper, accepted at the @mrl2024_emnlp workshop #EMNLP2024, explores this question! A 🧵:
ALT Language and Task Arithmetic with Parameter-Efficient Layers for Zero-Shot Summarization (arxiv.org/pdf/2311.09344)
Giorgos Vernikos retweeted
Can neural networks reason like pointcloud registration algorithms?
Excited to share our latest paper, "NAR-*ICP: Neural Execution of Classical ICP-based Pointcloud Registration Algorithms", with @PetarV_93, @d_d_martini, and @kula78!
arxiv.org/abs/2410.11031
More info below 🤖
Giorgos Vernikos retweeted
Can we achieve effective zero-shot summarization by combining language & task information from existing PEFT adapters with weight arithmetic?
Our paper, accepted at the @mrl2024_emnlp workshop #EMNLP2024, explores this question! A 🧵:
ALT Language and Task Arithmetic with Parameter-Efficient Layers for Zero-Shot Summarization (arxiv.org/pdf/2311.09344)
I am in Bangkok 🇹🇭 attending #ACL2024NLP! Tomorrow at 10:00 during Poster Session 4, I'll be presenting our work QE-fusion. If you're interested in translation, LLMs, and quality estimation, stop by for a chat!
Happy to share our latest preprint with @apopescubelis : "Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation"
📝preprint: arxiv.org/abs/2401.06688
Giorgos Vernikos retweeted
Congratulations to Dr. Katerina (@katemargatina) who passed her PhD viva today with no corrections!! 🎉🥳
Katerina has made very important contributions to active learning for NLP, including contrastive-based acquisition!
🚨 Excited to announce that our QE-fusion paper has been accepted to the #ACL2024NLP main conference! 🎉 See you all in Bangkok!! 🇹🇭
📰 camera-ready: arxiv.org/abs/2401.06688
Happy to share our latest preprint with @apopescubelis : "Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimation"
📝preprint: arxiv.org/abs/2401.06688
Giorgos Vernikos retweeted
#ICML2024 Paper 📣. tl;dr: a theoretical analysis of the minima of various contrastive learning losses + a novel loss family (seamless plugin) inspired by our theory, w/ robust & consistent downstream performance improvements. With @pakoromilas,
@tyiannak,@MihalisAN, Y. Panagakis
What do different contrastive learning (CL) losses actually optimize for? In our #ICML2024 paper, we provide a theoretical analysis and propose two loss functions that outperform conventional CL losses.
Full paper here: arxiv.org/abs/2405.18045 w/@gbouritsas
A thread 🧵
Giorgos Vernikos retweeted
LLMs become omnipresent, yet their deployment prevails in the cloud, threatening openness and user #privacy. As consumer devices get increasingly powerful, and smaller #open architectures are broadly available, a repo of #mobile-centric #LLMs is timely: github.com/stevelaskaridis/a…
